On Tue, 10 Jan 2023 at 14:22, <language.lawyer_at_[hidden]> wrote:

> > IOW, the example is not related to
> https://timsong-cpp.github.io/cppwp/n4868/over.ics.rank#3.3, you need
> multiple `f`s to start to care about that bullet.
>
> Actually, this should not be new to you, here is your 10 y.o. question
> https://stackoverflow.com/questions/11555950/distinguishing-between-user-defined-conversion-sequences-by-the-initial-standard


Amazing, thanks. It clearly didn't sink in 10 years ago!

Perhaps it would be helpful if https://wg21.link/over.ics.rank#note-1 were
to mention that these bullets cannot be used to rank ICSs but only apply to
the later provisions of https://wg21.link/over.match.best#general-2

I still wonder about "or they initialize the same class in an aggregate
initialization" in [over.ics.rank]/3.3, since it seems clear that in such
case the second SCS is always the identity conversion. Do you have any
thoughts about this part?